sidchaudhary/gtm-skills

88 skills

• account-plan

  Builds the engagement plan for one named strategic account: a buying committee map that names the roles still missing, a multi-threading assessment against the account's realistic committee size, and a 90-day action plan with an owner per play. Researches the account from public sources before asking for anything the web can answer. Use when a deal matters enough to plan deliberately, when it is single-threaded on one contact, or when a committee has gone quiet. Boundary: `opportunity-scoring` scores an existing deal's health and intent, while this skill builds the relationship plan for working the account. For one specific upcoming conversation use `call-preparation`.

  Self-containedActs undeclared2,534 words
• call-preparation

  Two modes for one specific meeting: pre-call prep, covering the agenda, discovery questions and the single thing that must be established, and post-call coaching, covering talk ratio, BANT or MEDDIC completeness, every objection raised and how it was handled, and a drill for the weakest habit. Use before a scheduled sales call, or immediately after one while the detail is fresh. Boundary: `call-notes` extracts deal facts from the same call for the follow-up and the CRM, while this skill grades the rep's own performance. `price-negotiation` covers a pricing or terms conversation specifically.

  Self-containedActs undeclared1,596 words
• objection-handling

  Takes a product description, ICP and target persona and outputs the five most likely objections, each with a specific acknowledgment, a response, and the follow-up question that moves past it. Use when preparing for cold calls, onboarding a new rep, or building objection handling into a sequence before it launches. Boundary: this prepares responses in advance for a persona. `inbox-management` handles an objection that has already arrived in a real reply, and `price-negotiation` handles a live pricing or terms push on one deal.

  Self-containedInstructions only1,917 words
• opportunity-scoring

  Scores one deal in depth on two independent axes, health and buyer intent, tracks the direction of travel since the last review, checks MEDDIC or BANT completeness, and returns a prioritised list of three to five specific next steps from where the deal lands. Use when a single deal needs an honest read before a forecast call, a renewal conversation, or a decision to keep investing in it. Boundary: `pipeline-review` triages the whole pipeline to decide which deals deserve this level of attention. For renewal risk on an existing customer use `opportunity-scoring`.

  Self-containedInstructions only2,229 words
• pipeline-review

  Reads a whole pipeline export and returns only what genuinely needs attention: deal velocity against your own stage norms, deals stuck past their stage median, risk signals, and where the forecast is most likely to slip. Lists only the deals that warrant action rather than padding to a fixed count. Use for a weekly pipeline review, before a forecast call, or when the pipeline number looks fine but the deals feel soft. Boundary: `opportunity-scoring` scores one deal in depth, while this skill triages across all of them to decide which ones deserve that.

  Self-containedInstructions only2,099 words
• price-negotiation

  Preps the negotiation strategy for one specific deal - anchor, concession ladder, walk-away point, and the questions that surface what the other side actually needs. Use before a pricing or terms conversation, not after it's already gone sideways. Pairs with opportunity-scoring and objection-handling.

  Self-containedInstructions only1,478 words
• win-loss-analysis

  Analyzes a batch of closed-won and closed-lost deals to find the real, evidence-backed reasons deals are actually won or lost, ranked by frequency, not a gut-feel retro. Use when the user wants to know why deals are actually closing or dying, not just track that they did. Pairs with objection-handling and competitive-analysis.

  Self-containedActs undeclared2,587 words
• creative-brief

  Turns a product and audience into a creative brief: distinct messaging angles selected from a fixed vault of fourteen, each mapped to a funnel stage and a channel, with placement specs and the reason that angle suits that stage. Use when starting a new ad campaign, when existing creative has fatigued and every variant is saying the same thing, or when a brief is needed before any asset gets designed. Boundary: this decides what the creative says and where it runs. `product-photography` then directs the photography and `onboarding-video` produces video. For the wording of a specific LinkedIn opening use `hook-writer`.

  Self-containedInstructions only2,257 words
• hook-writer

  Takes a LinkedIn post draft and generates four scored hook variants for the opening one or two lines before the see-more cutoff, each built on a distinct psychological framework, with a recommendation and the reason it wins. Use when a post's opening feels flat, or when testing hooks before publishing. Boundary: LinkedIn post openings only. `cold-email` does the equivalent job for cold email subject lines, and `cold-email` writes the opening line of a cold email body.

  Self-containedInstructions only1,491 words
• onboarding-video

  Produces a short conversion-focused onboarding video in Remotion for an iOS app, Android app or website, embedded into an onboarding flow to move New Customers toward their activation moment, with each beat mapped to a conversion milestone that can be tracked. Use when asked to create, build or generate an onboarding video, app preview, activation demo, or any short video that demonstrates a product feature to drive activation. Boundary: `product-photography` directs still photography and `creative-brief` produces the messaging brief. This skill renders an actual video.

  Self-containedActs undeclared4,175 words
• product-photography

  Turns a product into photography direction as a shootable set: subject, styling, environment, lighting, composition and camera treatment composed from a block system, scoped to five to eight images per product and matched to the store's existing catalogue treatment. Use when briefing a photographer or an image generator, before a product shoot, or when catalogue imagery is visually inconsistent from product to product. Boundary: `creative-brief` decides the messaging angle and placement before the shoot, and `product-page-optimization` reviews a live product page including its images. This skill only directs new imagery.

  Self-containedInstructions only2,626 words
• tone-of-voice

  Analyses real content samples and extracts a brand voice profile as a checkable artifact: tone dimensions each carrying at least one followable rule, required and banned vocabulary, sentence and paragraph patterns, channel adaptations, and a weighted scorecard with a publish-ready threshold to grade drafts against. Use when onboarding a writer or agency, when AI-generated copy is drifting off-brand, or when nobody on the team can say concretely what on-brand means. Boundary: `.agents/product-context.md` carries a short voice summary, while this skill produces the full profile and the scorecard.

  Self-containedActs undeclared2,703 words
• anomaly-detection

  Takes a time series of one metric and flags which recent points are genuinely outside its normal range, using a stated trailing-average-and-deviation method, not a gut read of a chart. Use when the user has a week's or month's worth of numbers for a metric and wants to know if something in it is actually unusual. Boundary: this skill flags anomalies in data the user provides. For designing the dashboard that surfaces this metric in the first place, use `kpi-dashboard`.

  Self-containedInstructions only2,462 words
• benchmark-analysis

  Takes one of the user's own metrics and checks it against a stated benchmark source, returning a clear over/under read and what that gap actually means. Use when the user wants to know if a number (churn rate, CAC, conversion rate, NPS) is good or bad relative to a real reference point, not just the number in isolation. Boundary: this skill does not have a live connection to any benchmark database. It compares against whatever source the user supplies, or discloses plainly when it is using general public knowledge instead.

  Self-containedInstructions only2,305 words
• cohort-analysis

  Groups customers by acquisition period and tracks a retention or revenue metric across the periods that follow, returned as a cohort table. Use when the user wants to know whether newer customers are performing better or worse than older ones, not just a single blended average. Boundary: this skill builds the cohort table itself. For the dashboard that displays it alongside other KPIs, use `kpi-dashboard`. For comparing the result against outside industry numbers, use `benchmark-analysis`.

  Self-containedInstructions only2,854 words
• contribution-margin

  Builds a per-SKU or per-order contribution margin stack (CM1, CM2, CM3) from raw revenue, cost, fee, and ad spend inputs, so the user can see which products actually make money after every variable cost, not just after product cost. Use when ROAS looks fine but profit doesn't, before scaling spend on a product, or when deciding which SKUs are worth promoting. Boundary: this skill computes margin from numbers the user hands over right now. For designing the recurring dashboard that surfaces margin over time, use `kpi-dashboard`.

  Self-containedActs undeclared2,827 words
• ecommerce-returns

  Takes a structured returns or RMA export with reason codes tied to SKUs and finds which SKUs have a genuine return concentration problem, and which coded reason is the likely root cause. Use when a return rate is rising, a specific SKU returns far above the catalog average, or returns are logged as a cost line instead of mined for the fix behind them. Boundary: this skill works from structured, coded return reasons tied to SKUs. If the input is free-text reviews or support tickets with no reason-code field, say so: the coded-reason analysis here does not apply and the raw text needs reading rather than tallying.

  Self-containedActs undeclared2,895 words
• growth-strategy

  Produces a prioritised growth plan: a maturity read, the candidate levers scored on effort, risk and reward, channel priorities, a 70-20-10 allocation where there is ongoing capacity, and an explicit statement of what is being declined for now. Use for quarterly or annual growth planning, or when there are more good ideas than the team can actually run. Boundary: `conversion-funnel` diagnoses where one specific funnel loses people, and `ab-test` designs the test for a lever once it has been chosen.

  Self-containedActs undeclared2,437 words
• inventory-planning

  Turns a SKU-level inventory export and sales history into a stockout/overstock risk brief, using a stated days-of-cover method against sales velocity and lead time, not a gut read of a stock report. Use when deciding which SKUs are safe to promote, which to protect from a planned campaign, or which are quietly overstocked. Boundary: differs from `anomaly-detection`, which flags one metric's time series against its own trailing average. This skill scores SKU-level stockout/overstock risk from on-hand units, sales velocity, and lead time, not a single metric's history.

  Self-containedInstructions only2,913 words
• kpi-dashboard

  Designs a KPI dashboard: the metric list with an explicit formula per metric, the visualisation form chosen from the question each answers, alert thresholds, and a layout wireframe. Use when building or rebuilding a marketing, sales, exec, product or CS dashboard, or when an existing one is not being read. Boundary: this designs the dashboard and its thresholds. `anomaly-detection` then judges whether a specific movement is genuinely abnormal, and `weekly-report` writes the recurring narrative readout that sits on top.

  Self-containedInstructions only2,617 words
• shipping-cost-analysis

  Compares what a store charges for shipping against what shipping actually costs, banded by order value and zone, and tests the free-shipping threshold against the store's real order distribution. Use when fulfillment cost is rising, the free-shipping threshold has never been recalculated, or margin looks worse on small orders than large ones. Boundary: differs from `contribution-margin`, which builds full order-level contribution margin across every cost line. This skill isolates shipping recovery and threshold placement only, and doesn't touch COGS, fees, or ad spend.

  Self-containedActs undeclared2,289 words
• weekly-report

  Produces one weekly operating readout for a store or product from exported performance, traffic, lifecycle, inventory, and support summaries, instead of scattered dashboards. Use when the user wants a single weekly view of what changed, what likely caused it, and what to do next. Boundary: this skill summarizes across whatever data the user brings this week. For a deep read on one specific area (margin, cohorts, checkout, search), use the matching skill directly and feed its output in here as one of the week's inputs.

  Self-containedInstructions only2,050 words
• ab-test

  Designs a runnable experiment: the hypothesis stated so that it can actually fail, the variants, Bayesian allocation with Thompson sampling, guardrail metrics, a holdout, required sample size and duration, explicit exit criteria, and the validity threats that would invalidate the read. Use when planning an A/B or multi-armed test, or when a previous test produced a result nobody trusts. Boundary: `website-personalization` designs personalisation rules that deliberately serve different audiences different content with no winner ever declared, whereas this skill runs a test to find one.

  Self-containedInstructions only2,591 words
• checkout-optimization

  Audits a cart and checkout flow directly, using screenshots or a walkthrough plus policy details, to find specific friction points in forms, payment coverage, trust signals, and step count. Use when cart abandonment is high, checkout conversion is weak, or the team wants a pre-launch or pre-scale checkout review. Boundary: differs from `conversion-funnel`, which diagnoses funnel drop-off against conversion benchmarks; this is a direct UX audit of the checkout flow itself, not a benchmark comparison.

  Self-containedInstructions only2,597 words
• conversion-funnel

  Diagnoses where a funnel loses people: stage-by-stage conversion against benchmarks, the drop-offs ranked by the spread they represent rather than by the worst absolute number, likely root causes, and an ordered fix roadmap. Use when conversion is weak on a marketing, sales or product funnel and it is not yet clear which step is responsible. Boundary: locates and ranks the leak, while `ab-test` designs the test for the fix. For checkout specifically use `checkout-optimization`, and for post-signup activation use `onboarding-flow`.

  Self-containedActs undeclared2,685 words
• landing-page

  Designs a conversion-focused landing page: the copy framework matched to traffic source and awareness stage, section order, form length, proof placement, and prototype-ready HTML with Tailwind. Use for a lead capture page, a signup page or a launch page that does not exist yet. Boundary: builds a new page from scratch. `product-page-optimization` reviews an existing product detail page, and `checkout-optimization` reviews cart and checkout.

  Reaches the webActs undeclared3,024 words
• onboarding-flow

  Designs the post-signup activation path: what happens between signup and first real value, in what order, which step is the actual aha moment rather than a convenient proxy for it, and how drop-off at each step gets diagnosed and fixed. Use when users sign up but do not activate, or when onboarding has never been designed deliberately. Boundary: covers signup to first value for a product. `email-campaign` writes the welcome messaging that carries it, and `conversion-funnel` diagnoses drop-off across any funnel rather than the activation path specifically.

  Self-containedInstructions only2,943 words
• paid-media-audit

  Triages paid ad spend, using ROAS, CAC, and spend concentration across channels, campaigns, and audiences, to find where budget is being wasted rather than assuming the ad account is the whole problem. Use when CPA is rising, ROAS is falling, or the team wants to know which campaigns or audiences to cut before adding more budget. Boundary: pairs with `conversion-funnel`, which diagnoses drop-off at a specific page or funnel step; this works one level up, at the level of which channels, campaigns, and audiences are burning spend.

  Self-containedInstructions only2,201 words
• pricing-strategy

  Designs pricing and packaging: the value metric to charge on, tier structure, the price points themselves, and the timing and framing of an increase. Flags the case where an absence of price objections is evidence of being underpriced. Use when setting prices for the first time, restructuring plans, or deciding whether and how to raise them. Boundary: sets the structure. `price-negotiation` handles discounting on one live deal, and `contribution-margin` computes what a given price actually earns after every variable cost.

  Self-containedInstructions only2,030 words
• product-page-optimization

  Reviews an existing product detail page, using the page itself, reviews, and buyer questions, to find clarity, trust, proof, and objection gaps, then returns a prioritized edit brief. Use when a product page isn't converting or before sending more traffic to it. Boundary: differs from `landing-page`, which generates new landing pages as HTML/Tailwind; this reviews a page that already exists and returns an edit brief, it doesn't generate new page code.

  Self-containedActs undeclared2,424 words
• website-personalization

  Designs personalisation rules that map an audience to a content variant, with the eligibility condition, the fallback for everyone who matches nothing, and a per-experience measurement plan. Use for dynamic content on pages, in emails or in-app, once the segments exist. Boundary: this serves different audiences different content permanently, with no winner declared. `ab-test` runs a test to pick one winner instead. Segment definitions come from `customer-segmentation`.

  Self-containedInstructions only2,312 words
• call-notes

  Reads a raw sales call transcript and extracts what the follow-up needs: deal signals, objections raised, pain points the prospect actually confirmed rather than ones the rep suggested, a stakeholder map, and one specific recommended next action. Use after a discovery call, before writing the follow-up email or updating the CRM record. Boundary: extracts deal facts for the follow-up and the CRM, while `call-preparation` grades the rep's own performance on that same call.

  Self-containedActs undeclared1,598 words
• competitive-analysis

  Two modes. Mode A researches a competitor from public sources into a structured, comparable profile: positioning, pricing, strengths and weaknesses, and the competitive implications. Mode B converts a prospect's known tool stack into a call-ready displacement angle: the specific inefficiency that stack creates, two call questions, and an email hook. Use when you need deep competitor research, a displacement angle for a named prospect, or both. Boundary: `win-loss-analysis` finds which competitors actually beat you using your own closed deals, whereas this skill works from public sources and a stated stack.

  Self-containedActs undeclared2,840 words
• lead-management

  Designs the lead-to-opportunity layer between marketing and sales: MQL scoring model, routing rules, speed-to-lead SLAs, and lifecycle stage definitions. Use when leads aren't reaching sales fast enough, marketing and sales disagree on what counts as qualified, or handoff is undefined. Boundary: opportunity-scoring scores one opportunity; pipeline-review reports on the whole pipeline; this skill is the layer before either exists.

  Self-containedInstructions only2,185 words
• lead-routing

  Designs the actual assignment logic for an already-qualified lead, round-robin, territory, account-owner, or score-threshold, with explicit tie-break and fallback rules. Use when leads are qualified but nobody has decided, in writing, exactly who they go to next. Boundary: `lead-management` designs the scoring model and lifecycle stages that decide whether a lead is qualified in the first place. This skill is the layer after that: once a lead is qualified, who actually gets it.

  Self-containedActs undeclared2,289 words
• marketing-automation

  Designs a marketing or sales automation as a specification: trigger, entry conditions, actions, branching, wait logic, error handling, the scheduled data assertions that catch a silent integration failure where nothing errors but values are wrong, plus a named owner, an audit date and a retirement condition. Use when automating an internal handoff or a repetitive process, or when an existing automation is misfiring and nobody can say why. Boundary: `customer-journey` designs customer-facing lifecycle journeys with messaging in them, while this skill designs the operational plumbing. `lead-management` covers lead scoring and routing specifically.

  Self-containedActs undeclared2,930 words
• product-launch-checklist

  Runs a one-time go/no-go readiness check before a product or campaign launch, covering stock, page, tracking, lifecycle flows, support, and margin as a checklist the launch owner signs off against. Use when a launch, restock, or spend increase is about to go live and the cost of finding a gap after the fact is high. Boundary: `marketing-automation` designs ongoing marketing and sales automation that runs indefinitely; this skill is a single pre-launch checklist for one specific launch date, not an automation build.

  Self-containedInstructions only1,788 words
• sales-enablement

  Produces sales collateral: one-pagers, ROI calculators, proposal templates, and playbooks, mapped to buyer persona and deal stage. Use when a rep needs a specific asset to hand a prospect. Boundary: for objection-handling responses specifically, use objection-handling: that skill already owns the objection library in this pack. This skill covers every other enablement asset type.

  Self-containedActs undeclared2,414 words
• churn-reduction

  Designs the systemic churn machinery: the cancel flow and its offramps, dynamic save offers ranked by cost against the margin they protect, churn risk scoring, and dunning sequences for failed payments. Splits the churn number into voluntary and involuntary before designing anything, because the two need opposite fixes. Use when reducing churn systemically rather than winning back one gone-dark account. Boundary: `opportunity-scoring` reads risk on one named account, and `repeat-purchase-rate` addresses a missing second order rather than a cancellation.

  Self-containedInstructions only2,645 words
• customer-journey

  Designs a multi-channel customer journey: the node sequence, conditional branching, wait logic, holdouts, exit conditions, and a global per-contact message cap across every channel rather than a cap per channel. Settles step count and channel mix before any messaging, because those drive recovery rates far more than copy does. Use for automation flows, onboarding, retention and win-back journeys. Boundary: designs the structure, while `email-campaign` writes the actual messages in each step. For internal operational automation use `marketing-automation`.

  Self-containedActs undeclared2,231 words
• customer-segmentation

  Builds lifecycle segments with RFM scoring calibrated against your own customer distribution rather than absolute cutoffs, plus behavioural signals, explicit filter logic per segment, and the staleness rule that says when a segment must be recomputed. Use for audience segmentation and targeting, or when existing segments have stopped matching reality. Boundary: defines who is in each segment. `email-campaign` and `customer-journey` then decide what those segments receive. For churn risk on one named account use `opportunity-scoring`.

  Self-containedActs undeclared2,966 words
• email-campaign

  Designs a lifecycle campaign end to end and decides whether it runs once or runs forever: email sequences with subject variants and deliverability gates, SMS and push with carrier registration and consent gates, and the recurring version of either with a cadence matched to how fast its signal actually moves. Use for welcome, nurture, convert, retain, win-back and announce campaigns on any channel, and when deciding whether a campaign should become an always-on motion. Boundary: writes the messaging. `customer-journey` designs the multi-step journey structure it sits inside, and `customer-segmentation` defines the audience it targets.

  Reaches the webInstructions only3,113 words
• promotional-campaigns

  Builds a promotional campaign per social channel from what is actually working in the market for similar products, and measures whether a promotion that already ran added real profit or just pulled demand forward (baseline-versus-promo-versus-recovery). Use to design a promo grounded in live competitor and market activity, or to grade one after a sale ends, a promo calendar is about to repeat, discount codes are leaking, or revenue rose while profit stayed flat. Boundary: `pricing-strategy` (Experimentation Lead) designs standing pricing tiers and price points; this designs and measures time-boxed promotions.

  Self-containedInstructions only3,153 words
• referral-program

  Designs a customer referral or advocacy program - incentive structure, the moment it's offered, tiers, and anti-abuse rules - so growth comes from existing customers, not just new acquisition spend. Use when the user wants existing customers to actively bring in new ones, not just stay retained. Pairs with churn-reduction and customer-segmentation.

  Self-containedInstructions only2,536 words
• repeat-purchase-rate

  Reviews a customer's post-first-purchase flow, reorder prompts, replenishment timing, second-purchase incentives, and pinpoints exactly what's blocking the next order. Use when repeat purchase rate is flat or falling, or the user wants to know why customers aren't coming back for a second order. Boundary: `customer-journey` designs a new multi-channel journey from scratch; this skill audits the flows that already exist against a fixed set of reorder stages and names the specific gap, it doesn't build the journey. `customer-segmentation` segments the whole customer base by RFM across all stages, not just the post-first-purchase window.

  Self-containedInstructions only2,126 words
• ad-angles

  Writes five or six genuinely different advertising angles for a single fixed offer, each naming a different buyer, a different pain and a different promised outcome, so the delivery system has something specific to aim with instead of one message wearing six costumes. Use when launching an offer, or when testing has killed the angles you had. Boundary: `creative-brief` briefs from a fixed library of fourteen angles mapped to funnel stages and placements; this invents new ones for one offer. Feed it `value-proposition` output and sharpen each line with `value-proposition`.

  Self-containedActs undeclared3,391 words
• ad-copy

  Picks the right classic copywriting formula for the placement and writes to it, pain-first structures for cold feeds and story structures only where length is earned, then diagnoses copy that reads fine but never earns the click, usually a formula with one step missing. Use to draft primary text, or when an ad gets impressions and no clicks. Boundary: `responsive-search-ads` writes Google assets against hard character limits; this writes long-form social copy, and `hook-writer` scores opening lines only.

  Self-containedActs undeclared2,770 words
• ad-design

  Turns an approved angle into finished, on-brand ad images and stages them in the ad account's image library, because an angle that never becomes a file never runs, and aims for native and specific rather than polished. Use after angles are approved and before the campaign is built. Boundary: `product-photography` directs photography from a composable block system and `creative-brief` writes the brief; this produces and uploads the finished ad file. `onboarding-video` covers onboarding video, not paid static.

  Self-containedActs undeclared2,844 words
• brand-kit

  Reads a public website into a working brand kit, the offer and the proof and the voice and the colour and type, so ads can be written and designed without anyone writing a creative brief first. Use before building any creative, or when ads keep coming out looking like they belong to nobody. Boundary: `tone-of-voice` scores an existing body of writing across six dimensions from samples you supply; this extracts a whole kit from a live site in one pass, and writes nothing back to `.agents/product-context.md`.

  Self-containedActs undeclared2,850 words
• budget-reallocation

  Ranks every line on both platforms by what a conversion actually costs there, names which are donors and which are recipients, and models three transfer sizes with the projected conversions and blended return for each, including moving money between platforms rather than only within one. Proposes scenarios for approval and moves nothing. Use when the split between platforms was set by history rather than by evidence. Boundary: `scaling-facebook-ads` adds budget to one proven winner in small steps without resetting its learning, `paid-media-audit` finds waste without proposing where it should go instead, and `stockout-alerts` pauses for stock reasons; this shifts money between existing lines.

  Self-containedActs undeclared3,203 words
• cost-per-acquisition

  Takes results from both ad platforms at once and returns one ranked list of why acquisition cost moved, each cause carrying a severity and the evidence behind it, and including the read neither account can produce alone: whether the two are bidding into the same people and inflating each other. Use when acquisition cost climbed on both platforms and the reason is not obvious in either one. Boundary: `daily-ad-check` and `daily-ad-check` look only inside one paid-social account, `google-ads-troubleshooting` works one search account's serving levels in dependency order, and `paid-media-audit` triages waste across every channel without joining platforms; this one joins two and ranks causes.

  Self-containedActs undeclared2,946 words
• daily-ad-check

  The read-only morning pass over one ad account: last seven complete days against the prior seven, budget going nowhere, ads fading, tests starved of spend, tracking that broke, and the single next test worth running, capped at five findings and changing nothing. Use daily, or on any account somebody else manages as an accountability check. Boundary: `daily-sales-report` reads a whole store's orders, revenue and total spend; this looks only inside the ad account, and `facebook-ads-audit` is the weekly decision review.

  Self-containedActs undeclared3,248 words
• facebook-ads-audit

  Reads a Meta ad account the way an analyst with no account to keep would: results broken out by angle, an explicit caveat on platform-reported return and attribution, and the three decisions the numbers actually support, with gaps named rather than filled with optimism. Use weekly, or before any decision to scale, kill or rebuild. Boundary: `ppc-reporting` is the Google equivalent, `paid-media-audit` triages waste across every channel at once, and `weekly-report` writes the whole-business readout.

  Self-containedActs undeclared2,960 words
• facebook-ads-campaign

  Builds a whole Meta ads campaign as a paused draft for review, one objective, one broad ad set, a small daily test budget and one ad per approved angle, keeping the structure simple on purpose because small budgets die of too much structure rather than too little. Use after creatives are approved and it is time to go live. Boundary: `email-campaign` composes lifecycle campaigns on owned channels, `google-ads-change-plan` sequences edits to an account already running, and `product-launch-tracking` monitors afterwards.

  Self-containedActs undeclared2,534 words
• google-ads-change-plan

  Turns a pile of Google Ads findings into an ordered change plan that respects dependencies, tracking before bidding and query evidence before negatives, naming each entity, its current state, the evidence behind the move, a rollback, and how the result gets measured. Use after analysis is done and the next move has to be unambiguous. Boundary: `automation-review` reviews another proposal by trying to refute it; this sequences approved findings into reversible steps, and `facebook-ads-campaign` builds from nothing.

  Self-containedActs undeclared2,925 words
• google-ads-conversion-tracking

  Decides whether Google Ads conversion data can be trusted to optimise against: which actions are primary, whether two actions count one event, whether a page view is drowning a demo, and what account evidence simply cannot prove without seeing the site or the CRM. Use before any bid, budget or target recommendation. Boundary: `meta-pixel` does the equivalent for the Meta pixel and its server-side events, and the findings here feed `smart-bidding`.

  Self-containedActs undeclared2,251 words
• google-ads-quality-score

  Uses the three Quality Score component ratings, expected click-through and ad relevance and landing-page experience, to find the weak link behind a keyword worth keeping, then drafts the smallest credible test, treating the score as a diagnostic rather than a number to chase. Use when valuable keywords show below-average components. Boundary: `product-page-optimization` audits a product page for on-site conversion and `responsive-search-ads` rewrites the assets; this only names which of the three links is weakest.

  Self-containedActs undeclared2,469 words
• google-ads-review

  Compares two complete and equal Google Ads periods in the account timezone, names the few campaigns or ad groups driving the movement, and ranks the next checks without changing anything, holding a bad-looking week open until conversion delay has been accounted for. Use weekly. Boundary: `weekly-report` writes the whole-business readout across every channel and `daily-ad-check` is the daily paid-social pass, while `ppc-reporting` defines the numbers this one explains.

  Self-containedActs undeclared2,921 words
• google-ads-troubleshooting

  Finds why Google Ads serving or reporting changed, working the clue levels in dependency order, account status then campaign status reasons then disapprovals then budget then rank then destination then learning state, keeping what is observed separate from what is merely suspected. Use when impressions, clicks, spend or conversions fall without an obvious reason. Boundary: `conversion-funnel` hunts drop-off in an owned funnel and `product-catalog-audit` tracks shopping disapprovals over time; this diagnoses one account's serving now.

  Self-containedActs undeclared2,750 words
• keyword-expansion

  Discovers the adjacent, net-new search keywords worth owning from a proven query (autocomplete, related searches, the search-terms report, competitor ads), then gives each one - the proven query and the discovered terms - deliberate keyword, ad-group and landing-page coverage without creating duplicates or mixing intents, choosing a match type for the job and holding any proposal no page or ad can honestly answer. Use after a review finds real demand worth owning. Boundary: `search-term-report` classifies the queries and `negative-keywords` handles the losers, while `keyword-intent` decides which cluster a promoted query joins.

  Self-containedActs undeclared2,792 words
• keyword-intent

  Groups paid search queries by the answer each searcher actually needs, pricing or comparison or login or category browsing or education, then maps every cluster to one ad-group promise and one page that can honestly deliver it, calling out an intent the site cannot answer as a gap. Use before adding keywords or restructuring Search. Boundary: site search tunes on-site search ranking inside your own store, not paid search, and `keyword-expansion` routes proven queries into these clusters.

  Self-containedActs undeclared1,922 words
• meta-ad-library

  Reads a rival's live ads in the public Meta Ad Library and separates proven messages from noise using two signals, how many creative variations one message has and how long it has kept running, then turns the survivors into opportunities for your own offer. Use before writing angles, or when yours have all plateaued. Boundary: `competitive-analysis` profiles a rival's positioning, pricing and weak spots from their website; this reads only what they are paying to say right now, and feeds `ad-angles`.

  Self-containedActs undeclared2,883 words
• meta-audience-targeting

  Researches what targeting is actually left on Meta after interest categories were retired, sizes the first-party customer lists and lookalike seeds that remain, and returns an honest verdict that is often go broad and fix the message instead. Use before a launch when you hold a list worth seeding, or when delivery keeps landing in the wrong crowd. Boundary: `customer-segmentation` designs owned-audience stages for channels you control; this only decides what to tell an ad platform. When the verdict is a message problem it hands off to `ad-angles`.

  Self-containedActs undeclared2,341 words
• meta-pixel

  Sets up the Meta pixel and its conversions from nothing - base install mechanism, standard events mapped to the funnel with the right value/currency parameters, the Conversions API wired alongside the browser pixel with matching event_id deduplication, AEM priority, and a Test-Events verification - or audits an existing one for whether the pixel and its server-side events are telling the truth: events that quietly stopped firing, one purchase counted twice, deduplication keys that do not match, and attribution windows that flatter. Use to install tracking properly, before trusting any reported number, and before building catalog or retargeting work on top of it. Boundary: `google-ads-conversion-tracking` does the equivalent for Google conversion actions and goals, and `marketing-automation` designs the automation that fires events; this only audits what arrived.

  Self-containedActs undeclared3,163 words
• negative-keywords

  Turns confirmed bad-fit searches into an approval-ready negative keyword draft, choosing the narrowest useful scope and testing every proposed negative against protected searches and existing positive keywords, because negative match types do not behave like positive ones. Use after a search-term review, before excluding any traffic. Boundary: `search-term-report` produces the confirmed candidates this drafts from, and `keyword-expansion` handles the winners travelling the other way.

  Self-containedActs undeclared2,111 words
• ppc-reporting

  Turns a Google Ads account into five to seven numbers tied to the business outcome, each carrying its formula, its target and its caveat, starting from what the business needs rather than from whatever the platform puts on the dashboard. Use weekly for an owner update that has to fit one screen. Boundary: `facebook-ads-audit` is the paid-social equivalent and `google-ads-review` explains what moved between two periods; this decides which numbers get watched. `kpi-dashboard` designs whole-product dashboards instead.

  Self-containedActs undeclared2,970 words
• product-feed

  Sets up the product catalog and the product sets that let ads pull live inventory instead of static images, treating each set as its own promise: best sellers, under fifty, new arrivals. Use after tracking is verified, for stores with more SKUs than per-product creative can cover. Boundary: `product-catalog-audit` checks an outbound feed for disapprovals and attribute breakage; this builds the sets themselves. Requires `meta-pixel` to pass first, or it automates showing people the wrong products.

  Self-containedActs undeclared2,148 words
• responsive-search-ads

  Writes a complete Google responsive search ad from real query intent and claims the landing page can support, then validates every headline, description, path, sitelink and callout against its character limit, since each asset has to work alone and beside the others. Use when a focused ad group needs credible copy. Boundary: `ad-copy` writes long-form paid-social primary text with no character ceiling, and `cold-email` writes email subject lines.

  Self-containedActs undeclared2,763 words
• scaling-facebook-ads

  Drafts the budget rules that scale a proven ad without resetting its learning: increments of roughly twenty percent no more than once a day, spend caps, automatic pauses for what has proven it loses, and a schedule. Every rule is proposed for approval, never applied. Use when an angle has earned more budget and manual edits on instinct keep crashing it. Boundary: `stockout-alerts` pauses spend for stock reasons and `margin-monitoring` watches per-SKU profitability; this paces a winner.

  Self-containedActs undeclared3,102 words
• search-term-report

  Sorts the queries Google actually bought into keep, review and exclude candidates, using business fit first and performance evidence second, on the rule that a query with no conversions can be plain waste, an under-tested one, or simply a slow-converting offer. Use weekly or monthly, whenever the search-term report gets reviewed. Boundary: `value-proposition` mines customer language for copy rather than bought queries. Confirmed candidates go to `negative-keywords` and winners to `keyword-expansion`.

  Self-containedActs undeclared2,570 words
• smart-bidding

  Matches a Google Ads bidding approach to a trusted conversion goal, the observed volume, the conversion delay and the business target, then drafts the plan for judging it, changing one major variable at a time and reading the result only after the data matures. Use before changing strategy, target cost, target return or budget. Boundary: `google-ads-conversion-tracking` must pass first, because bidding will optimise whatever goal it is handed, and `scaling-facebook-ads` paces paid-social increments instead.

  Self-containedActs undeclared2,155 words
• value-proposition

  Mines real buyer language from reviews, forums, tickets and competitor ad comments, then turns the strongest pain into one specific promise line and stress-tests it against the one-second recognition test, on the rule that promises fail by generalising, since a promise to everyone reads to the delivery system as a promise to nobody. Use after mining a pain and before that pain becomes copy, or when an ad earns impressions but no clicks. Boundary: `value-proposition` supplies the raw pain and `ad-angles` turns sharpened lines into whole angles, while `hook-writer` scores social openings against psychological frameworks instead.

  Self-containedActs undeclared3,452 words
• appointment-setting

  Turns a warm conversation into a booked meeting and keeps it there - the booking message, the reschedule, the confirmation, the day-before reminder, and, when a booked meeting is missed, the 3-stage no-show recovery (hour-one, day-two, one-week close) plus a history-based stop rule. Use when a prospect is ready to talk and the user wants to lock a time without back-and-forth, or to recover a no-show without sounding annoyed. Pairs with cold-email and account-plan; absorbs the former standalone missed-meeting-email (there is no separate no-show skill).

  Self-containedInstructions only1,772 words
• cold-email

  Takes prospect research, a specific trigger signal, and a product value prop, and writes a complete cold email body under 120 words, personalized to the signal. Use when the user wants to write a cold email to a specific prospect using their research inputs. Pairs with account-plan, lead-list, and cold-email.

  Self-containedInstructions only2,978 words
• email-sequence

  Builds AND audits email sequences for any business type - cold outbound for B2B/SaaS, and lifecycle (welcome, abandoned-cart, browse-abandon, post-purchase, win-back) for ecommerce or any website. Build mode produces a full multi-touch sequence with a unique angle at every step, drawn from the brand kit and ICP; audit mode runs a gap analysis per email, rewrites the weakest, and recommends missing steps. Use to create a sequence from scratch, or when reply rates are low and the cause is unclear. Boundary: `cold-email` writes one first-touch email from scratch; this builds or fixes the whole multi-touch sequence.

  Self-containedInstructions only2,502 words
• inbox-management

  Sorts a batch of inbound sales replies into Interested, Later, Referred, Objection, Dead, or Angry, with the evidence and next action for each. Use when the user has replies piling up and needs to know what to do with each one, not just what it says. Pairs with inbox-management.

  Self-containedInstructions only3,197 words
• intent-data

  Takes a batch of intent signals from across the pipeline and returns a ranked outreach list with the signal, the rationale, and a suggested channel per account, ordered so the week starts with action rather than reading. Use when prioritising a week of outreach around what actually changed. Boundary: ranks by fresh signal and its decay, whereas `lead-scoring` ranks by static ICP fit. `cold-email` then writes to whatever this surfaces.

  Self-containedInstructions only2,352 words
• lead-list

  Builds and qualifies a prospect list from an ICP definition: sourced from data the user provides or from public web research, scored Hot/Warm/Cold/Skip with evidence and confidence per lead. Use when the user wants to build a target account or lead list before outreach. Pairs with lead-scoring (scores a list that already exists) and feeds cold-email.

  Self-containedActs undeclared2,527 words
• lead-scoring

  Scores an account list that already exists against ICP criteria and returns a priority tier of High, Medium or Low with a one-sentence outreach rationale per account, plus what was inferred rather than observed. Use when prioritising a list of accounts before an outbound campaign. Boundary: `lead-list` sources and qualifies a list that does not exist yet, whereas this skill scores one you already have. `list-cleaning` fixes data quality before either runs.

  Self-containedInstructions only2,215 words
• list-cleaning

  Cleans a raw prospect list before it goes into a sequence - dedupes, flags wrong titles, stale roles, wrong companies, and broken data - without silently deleting anything. Use when the user has a raw export, scrape, or CRM pull that needs hygiene before use. Pairs with lead-list and lead-scoring.

  Self-containedInstructions only2,201 words
• automation-design

  Turns a recurring store question into a runnable loop specification: a cadence, an objective gate that can fail the work, a stop condition, an approval boundary, and the state file it reads. Use when the user keeps re-running the same store check by hand and wants it to run on a schedule instead. Boundary: this designs a loop that does not exist yet, and its defining output is the gate. `automation-ledger` creates and maintains the state file that loops read and append to once they are running. `marketing-automation` designs in-platform marketing and sales automation rather than an agent-side loop. Boundary: `automation-ledger` creates and maintains the shared state file that loops read and write; this one writes the loop specification itself.

  Self-containedInstructions only1,815 words
• automation-ledger

  Creates and maintains `.agents/store-loop-ledger.md`, the state file every store loop reads and appends to, recording what each run checked, what it flagged, what changed, and which recurring patterns to stop flagging. Use before running any loop on a cadence, and whenever a loop keeps re-reporting something already dismissed. Boundary: `automation-design` specifies a new loop and its gate, whereas this skill owns the shared memory that existing loops write to between runs. `.agents/product-context.md` stores who the business is and rarely changes; this file stores what the loops have learned and changes every run. Boundary: `automation-design` writes the loop specification; this one only creates and maintains `.agents/store-loop-ledger.md`, the state file those loops read from and write back to.

  Self-containedInstructions only1,678 words
• automation-review

  Reviews another loop's proposed change before it reaches a human, trying to refute it rather than confirm it - checking sample size, input freshness, whether the gate could actually have failed, and whether this exact change was already tried and reverted. Use as the second half of every loop that proposes an action. Boundary: `automation-design` specifies a loop before it runs. This skill reviews one specific proposal a loop already produced, and its default verdict is reject.

  Self-containedInstructions only1,795 words
• daily-sales-report

  Runs the daily pass over a store's orders, revenue, and ad spend, flagging only what moved outside its normal band against a trailing baseline, so the morning read is a short ranked list instead of three dashboards. Use as the first loop any store installs. Boundary: `weekly-report` writes one weekly narrative readout across whatever data the user brings. This loop runs daily, diffs against a stored baseline, and reports only exceptions - it is not a summary of everything.

  Self-containedInstructions only2,230 words
• margin-monitoring

  Watches per-SKU contribution margin on a recurring cadence and flags the SKUs that crossed from profitable to unprofitable since the last run, separating the ones killed by ad spend from the ones that were already losing money. Use weekly, or after any price, supplier, or shipping change. Boundary: `contribution-margin` computes the margin stack once from numbers handed over now. This loop runs that stack repeatedly, diffs it against the last run, and reports crossings rather than levels.

  Self-containedInstructions only2,173 words
• product-catalog-audit

  Audits a whole product catalog export for two things at once: content completeness (missing attributes, thin or empty descriptions, duplicate copy, image gaps) and feed / channel deliverability (disapproval-risk attributes, feed-versus-live-page price and availability mismatches, category mapping, variant-ID confusion), ranked by revenue rather than row count. Runs in either of two modes: a one-time FULL audit of the whole catalogue, or a recurring DELTA watch that reports only what broke since the last run. Use when the catalog has too many SKUs to review page by page, before a marketplace or feed push, when only the best sellers seem to have real content, or on a daily/every-other-day cadence to catch a fresh disapproval before it is buried under a standing backlog. Boundary: this is the single catalog-health skill and it absorbs the former standalone feed watch; there is no separate shopping-feed skill.

  Self-containedInstructions only2,368 words
• product-launch-tracking

  Watches a new product's first weeks against pre-set early-signal thresholds and a fixed test budget, proposing scale or stop before the spend runs away. Use for the fixed window after a product launch or restock. Boundary: `product-launch-checklist` is the one-time go/no-go checklist run before launch day. This loop runs after launch, on a cadence, for a bounded window, and its gate is early performance rather than readiness.

  Self-containedInstructions only2,608 words
• stockout-alerts

  Cross-checks active ad spend against on-hand inventory on a daily cadence and proposes pausing spend on products that cannot be shipped or cannot cover their own acquisition cost. Use daily on any store running paid traffic to a catalog that moves. Boundary: `inventory-planning` scores stockout and overstock risk across the catalog for planning. This loop only looks at the intersection of low stock and live spend, and its output is a pause list.

  Self-containedInstructions only1,816 words